Convolutional Neural Networks¶
Project: Write an Algorithm for a Dog Identification App¶
In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the Jupyter Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.
The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this Jupyter notebook.
Why We're Here¶
In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).
In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!
The Road Ahead¶
We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.
- Step 0: Import Datasets
- Step 1: Detect Humans
- Step 2: Detect Dogs
- Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
- Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 5: Write your Algorithm
- Step 6: Test Your Algorithm
Step 0: Import Datasets¶
Make sure that you've downloaded the required human and dog datasets:
Note: if you are using the Udacity workspace, you DO NOT need to re-download these - they can be found in the /data folder as noted in the cell below.
Download the dog dataset. Unzip the folder and place it in this project's home directory, at the location
/dog_images.Download the human dataset. Unzip the folder and place it in the home directory, at location
/lfw.
Note: If you are using a Windows machine, you are encouraged to use 7zip to extract the folder.
In the code cell below, we save the file paths for both the human (LFW) dataset and dog dataset in the numpy arrays human_files and dog_files.
import numpy as np
from glob import glob
# load filenames for human and dog images
human_files = np.array(glob("/data/lfw/*/*"))
dog_files = np.array(glob("/data/dog_images/*/*/*"))
# print number of images in each dataset
print('There are %d total human images.' % len(human_files))
print('There are %d total dog images.' % len(dog_files))
Step 1: Detect Humans¶
In this section, we use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images.
OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory. In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(human_files[0])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(faces))
# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.
In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.
Write a Human Face Detector¶
We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
(IMPLEMENTATION) Assess the Human Face Detector¶
Question 1: Use the code cell below to test the performance of the face_detector function.
- What percentage of the first 100 images in
human_fileshave a detected human face? - What percentage of the first 100 images in
dog_fileshave a detected human face?
Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.
Answer: (You can print out your results and/or write your percentages in this cell)
from tqdm import tqdm
human_files_short = human_files[:100]
dog_files_short = dog_files[:100]
#-#-# Do NOT modify the code above this line. #-#-#
## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
human_in_human_img = []
dog_in_human_img = []
for filename in tqdm(human_files_short):
if face_detector(filename) is True:
human_in_human_img.append(filename)
for filename in tqdm(dog_files_short):
if face_detector(filename) is True:
dog_in_human_img.append(filename)
print(f'Percentage of first 100 images in human_files with a human face = {len(human_in_human_img)}%')
print(f'Percentage of first 100 images in dog_files with a human face = {len(dog_in_human_img)}%')
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.
Assess False Positive Results¶
for face in dog_in_human_img:
img = cv2.imread(face)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
print('Number of faces detected:', len(faces))
W, H, _ = img.shape
for (x,y,w,h) in faces:
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
print(f'x = {x}, y = {y}, w = {w}, h = {h}, w/W = {100*w/W:.1f}%, h/H = {100*h/H:.1f}%')
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
- The cell above shows the images where a human face has been detected in the images of dogs.
- In some cases there are indeed humans in the image.
- However in some other cases, the face detector has detected a face in a region of the image where there isn't a face
- Sometimes, these false positive faces images are in small regions of the image.
- A rule could be set where if the width or height of the bounding box must be sufficiently big in comparison to the dimensions of the image
- The human images could be analysed to assess the relative size of the bounding boxes, then the threshold could be set based on this.
Assess False Negative Results¶
for face in set(human_files_short) - set(human_in_human_img):
img = cv2.imread(face)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
- Only 2 images had false negative results which is low
- Maybe the model couldn't recognise the face due to the presence of sunglasses or the facial expression
### (Optional)
### TODO: Test performance of anotherface detection algorithm.
### Feel free to use as many code cells as needed.
Step 2: Detect Dogs¶
In this section, we use a pre-trained model to detect dogs in images.
Obtain Pre-trained VGG-16 Model¶
The code cell below downloads the VGG-16 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories.
import torch
import torchvision.models as models
# define VGG16 model
VGG16 = models.vgg16(pretrained=True)
# check if CUDA is available
use_cuda = torch.cuda.is_available()
# move model to GPU if CUDA is available
if use_cuda:
VGG16 = VGG16.cuda()
Given an image, this pre-trained VGG-16 model returns a prediction (derived from the 1000 possible categories in ImageNet) for the object that is contained in the image.
(IMPLEMENTATION) Making Predictions with a Pre-trained Model¶
In the next code cell, you will write a function that accepts a path to an image (such as 'dogImages/train/001.Affenpinscher/Affenpinscher_00001.jpg') as input and returns the index corresponding to the ImageNet class that is predicted by the pre-trained VGG-16 model. The output should always be an integer between 0 and 999, inclusive.
Before writing the function, make sure that you take the time to learn how to appropriately pre-process tensors for pre-trained models in the PyTorch documentation.
VGG16.eval()
from PIL import Image
import torchvision.transforms as transforms
from torch.autograd import Variable
# VGG-16 Takes 224x224 images as input, so we resize all of them
def image_loader(image_name, imsize=224):
"""
load single image, returns tensor
Ref: https://discuss.pytorch.org/t/how-to-classify-single-image-using-loaded-net/1411
"""
loader = transforms.Compose([
transforms.Resize(255),
transforms.CenterCrop(imsize),
# transforms.RandomResizedCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])])
image = Image.open(image_name)
image = loader(image).float()
image = Variable(image, requires_grad=True)
image = image.unsqueeze(0) #this is for VGG, may not be needed for ResNet
return image
def VGG16_predict(img_path):
'''
Use pre-trained VGG-16 model to obtain index corresponding to
predicted ImageNet class for image at specified path
Args:
img_path: path to an image
Returns:
Index corresponding to VGG-16 model's prediction
'''
## TODO: Complete the function.
## Load and pre-process an image from the given img_path
## Return the *index* of the predicted class for that image
image = image_loader(img_path)
if use_cuda:
image = image.cuda()
output = VGG16(image)
_, preds_tensor = torch.max(output, 1)
preds = np.squeeze(preds_tensor)
return preds
test_dog_image = dog_files[0]
test_dog_image
img = cv2.imread(test_dog_image)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
pred = VGG16_predict(test_dog_image)
if use_cuda:
pred = pred.cpu()
pred.numpy() == 243
(IMPLEMENTATION) Write a Dog Detector¶
While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained VGG-16 model, we need only check if the pre-trained model predicts an index between 151 and 268 (inclusive).
Use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
## TODO: Complete the function.
pred = VGG16_predict(img_path)
if use_cuda:
pred = pred.cpu()
if pred.numpy() in list(range(151, 268+1)):
return True
else:
return False
dog_detector(test_dog_image)
(IMPLEMENTATION) Assess the Dog Detector¶
Question 2: Use the code cell below to test the performance of your dog_detector function.
- What percentage of the images in
human_files_shorthave a detected dog? - What percentage of the images in
dog_files_shorthave a detected dog?
Answer:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
human_in_dog_img = []
dog_in_dog_img = []
for filename in tqdm(human_files_short):
if dog_detector(filename) is True:
human_in_dog_img.append(filename)
for filename in tqdm(dog_files_short):
if dog_detector(filename) is True:
dog_in_dog_img.append(filename)
print(f'Percentage of first 100 images in human_files with a dog = {len(human_in_dog_img)}%')
print(f'Percentage of first 100 images in dog_files with a dog = {len(dog_in_dog_img)}%')
for face in human_in_dog_img:
img = cv2.imread(face)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
for face in set(dog_files_short) - set(dog_in_dog_img):
img = cv2.imread(face)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
We suggest VGG-16 as a potential network to detect dog images in your algorithm, but you are free to explore other pre-trained networks (such as Inception-v3, ResNet-50, etc). Please use the code cell below to test other pre-trained PyTorch models. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.
### (Optional)
### TODO: Report the performance of another pre-trained network.
### Feel free to use as many code cells as needed.
Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶
Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 10%. In Step 4 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.
We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have trouble distinguishing between a Brittany and a Welsh Springer Spaniel.
| Brittany | Welsh Springer Spaniel |
|---|---|
 |
 |
It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).
| Curly-Coated Retriever | American Water Spaniel |
|---|---|
 |
 |
Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
 |
 |
 |
We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.
Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!
(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset¶
Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dog_images/train, dog_images/valid, and dog_images/test, respectively). You may find this documentation on custom datasets to be a useful resource. If you are interested in augmenting your training and/or validation data, check out the wide variety of transforms!
import torch
import torchvision.models as models
from PIL import Image
import torchvision.transforms as transforms
from torch.autograd import Variable
import os
from torchvision import datasets
### TODO: Write data loaders for training, validation, and test sets
## Specify appropriate transforms, and batch_sizes
# check if CUDA is available
use_cuda = torch.cuda.is_available()
imsize = 224
train_loader = transforms.Compose([
transforms.Resize(255),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(10),
transforms.CenterCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.5, 0.5, 0.5],
std=[0.5, 0.5, 0.5]
)
])
test_loader = transforms.Compose([
transforms.Resize(255),
transforms.CenterCrop(imsize),
# transforms.RandomResizedCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.5, 0.5, 0.5],
std=[0.5, 0.5, 0.5]
)
])
trainset = datasets.ImageFolder('/data/dog_images/train/', transform=train_loader)
validset = datasets.ImageFolder('/data/dog_images/valid/', transform=test_loader)
testset = datasets.ImageFolder('/data/dog_images/test/', transform=test_loader)
loaders_scratch = {}
loaders_scratch['train'] = torch.utils.data.DataLoader(trainset, batch_size=32, shuffle=True)
loaders_scratch['valid'] = torch.utils.data.DataLoader(validset, batch_size=32, shuffle=True)
loaders_scratch['test'] = torch.utils.data.DataLoader(testset, batch_size=32, shuffle=True)
Question 3: Describe your chosen procedure for preprocessing the data.
- How does your code resize the images (by cropping, stretching, etc)? What size did you pick for the input tensor, and why?
- Did you decide to augment the dataset? If so, how (through translations, flips, rotations, etc)? If not, why not?
Answer:
- The dog images are transformed to have final pixel size of 224 x 244 pixels, this is done using the following 3 steps:
- First the image is sized to 255 x 255 pixels using
transforms.Resize()which scales the width and height of the image. This does change the aspect ratio of the image, therefore if an image is very tall or wide it would become severely distorted and effect the performance of the model - The training images only are randomly flipped (
transforms.RandomHorizontalFlip()) and rotatedtransforms.RandomRotation(10)up to an angle of 10 degrees to give some additional variation in the position of the dogs in the image. This should help the model generalise to other unseen images. - Then a region of 224 x 224 pixels is cropped from the center of the image (This is consistent with the sizes of the images in the ImageNet database which will be needed during the transfer learning stage of this project.)
- First the image is sized to 255 x 255 pixels using
Visualize a Batch of Training Data¶
import matplotlib.pyplot as plt
%matplotlib inline
# helper function to un-normalize and display an image
def imshow(img):
img = img / 2 + 0.5 # unnormalize
plt.imshow(np.transpose(img, (1, 2, 0))) # convert from Tensor image
# specify the image classes
# classes = ['airplane', 'automobile', 'bird', 'cat', 'deer',
# 'dog', 'frog', 'horse', 'ship', 'truck']
# obtain one batch of training images
dataiter = iter(loaders_scratch['train'])
images, labels = dataiter.next()
images = images.numpy() # convert images to numpy for display
# plot the images in the batch, along with the corresponding labels
fig = plt.figure(figsize=(25, 4))
# display 20 images
for idx in np.arange(20):
ax = fig.add_subplot(2, 20/2, idx+1, xticks=[], yticks=[])
imshow(images[idx])
ax.set_title(labels[idx]) # ax.set_title(classes[labels[idx]])
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. Use the template in the code cell below.
import torch.nn as nn
import torch.nn.functional as F
# define the CNN architecture
class Net(nn.Module):
### TODO: choose an architecture, and complete the class
def __init__(self):
super(Net, self).__init__()
## Define layers of a CNN
# convolutional layer
self.conv1 = nn.Conv2d(3, 16, 3, padding=1)
# W=224, F=3, P=1, S=1, (W−F+2P)/S+1=224
# max pooling layer
self.pool2 = nn.MaxPool2d(2, 2)
# 224/2=112
# convolutional layer
self.conv2 = nn.Conv2d(16, 32, 3, padding=2)
# W=112, F=3, P=2, S=1, (W−F+2P)/S+1=114
# max pooling layer
self.pool3 = nn.MaxPool2d(3, 3)
# 114/3=38
# convolutional layer
self.conv3 = nn.Conv2d(32, 64, 3, padding=0)
# W=38, F=3, P=0, S=1, (W−F+2P)/S+1=36
# max pooling layer
# self.pool3 = nn.MaxPool2d(3, 3)
# 36/3=12
self.dropout = nn.Dropout(0.2)
# fully connected layer
self.fc1 = nn.Linear(64 * 12 * 12, 700)
self.fc2 = nn.Linear(700, 300)
self.fc3 = nn.Linear(300, 133)
def forward(self, x):
## Define forward behavior
x = self.pool2(F.relu(self.conv1(x)))
x = self.pool3(F.relu(self.conv2(x)))
x = self.pool3(F.relu(self.conv3(x)))
x = x.view(-1, 64 * 12 * 12)
x = self.dropout(x)
x = F.relu(self.fc1(x))
x = self.dropout(x)
x = F.relu(self.fc2(x))
x = self.dropout(x)
x = self.fc3(x)
return x
#-#-# You so NOT have to modify the code below this line. #-#-#
# instantiate the CNN
model_scratch = Net()
# move tensors to GPU if CUDA is available
if use_cuda:
model_scratch.cuda()
Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step.
Answer:
print(model_scratch)
Model Architecture
| Name | Type | in_channels | out_channels | kernel_size | stride | padding | p | Size | Depth |
|---|---|---|---|---|---|---|---|---|---|
| conv1 | Conv2d | 3 | 16 | (3,3) | (1,1) | (1,1) | - | 224 x 224 | 16 |
| - | ReLU | 16 | 16 | - | - | - | - | 224 x 224 | 16 |
| pool2 | MaxPool2d | 16 | 16 | (2,2) | (2,2) | (0,0) | - | 112 x 112 | 16 |
| conv1 | Conv2d | 16 | 32 | (3,3) | (1,1) | (2,2) | - | 114 x 114 | 32 |
| - | ReLU | 32 | 32 | - | - | - | - | 114 x 114 | 32 |
| pool3 | MaxPool2d | 32 | 32 | (3,3) | (3,3) | (0,0) | - | 38 x 38 | 32 |
| conv3 | Conv2d | 32 | 64 | (3,3) | (1,1) | (0,0) | - | 36 x 36 | 64 |
| - | ReLU | 64 | 64 | - | - | - | - | 36 x 36 | 64 |
| pool3 | MaxPool2d | 64 | 64 | (3,3) | (3,3) | (0,0) | - | 12 x 12 | 64 |
| - | Flatten | 64 x 12 x 12 | 9216 | - | - | - | - | 9216 | 1 |
| dropout | Dropout | - | - | - | - | - | 0.2 | 9216 | 1 |
| fc1 | Linear | 9216 | 700 | - | - | - | - | 700 | 1 |
| - | ReLU | 700 | 700 | - | - | - | - | 700 | 1 |
| dropout | Dropout | - | - | - | - | - | 0.2 | 700 | 1 |
| fc2 | Linear | 700 | 300 | - | - | - | - | 300 | 1 |
| - | ReLU | 300 | 300 | - | - | - | - | 300 | 1 |
| dropout | Dropout | - | - | - | - | - | 0.2 | 300 | 1 |
| fc3 | Linear | 300 | 133 | - | - | - | - | 133 | 1 |
Details of Model
- The CNN model consists of 3 convolutional layers which are each followed by ReLU activation functions
- The small kernel sizes of the first convolutional and maxpool layers (3 x 3 and 2 x 2 respectively) allows the model to learn about details in the image such as the dogs' fur
- Then a further 2 maxpool layers reduces the dimensionality of the layer to 12 x 12
- The channels in the convolutional layers were increased up to 64
- The final maxpool layer is flattened to created fully connected layer with 9,216 nodes
- The model has a further two fully connected layers, where the final layer has 133 nodes which are the same as the number of classes
- In the fully connected layers, 20% of the hidden nodes are removed for each training pass to ensure the model doesn't become too biased towards certain hidden nodes and the the model generalizes well
(IMPLEMENTATION) Specify Loss Function and Optimizer¶
Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_scratch, and the optimizer as optimizer_scratch below.
import torch.optim as optim
### TODO: select loss function
criterion_scratch = nn.CrossEntropyLoss()
### TODO: select optimizer
optimizer_scratch = optim.Adam(model_scratch.parameters(), lr=0.001)
(IMPLEMENTATION) Train and Validate the Model¶
Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_scratch.pt'.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
from datetime import datetime
def train(n_epochs, loaders, model, optimizer, criterion, use_cuda, save_path):
"""returns trained model"""
# initialize tracker for minimum validation loss
valid_loss_min = np.Inf
start_datetime = datetime.now()
for epoch in range(1, n_epochs+1):
# initialize variables to monitor training and validation loss
train_loss = 0.0
valid_loss = 0.0
###################
# train the model #
###################
model.train()
for batch_idx, (data, target) in enumerate(loaders['train']):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
## find the loss and update the model parameters accordingly
## record the average training loss, using something like
## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
# clear the gradients of all optimized variables
optimizer.zero_grad()
# forward pass: compute predicted outputs by passing inputs to the model
output = model(data)
# calculate the batch loss
loss = criterion(output, target)
# backward pass: compute gradient of the loss with respect to model parameters
loss.backward()
# perform a single optimization step (parameter update)
optimizer.step()
# update training loss
train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
print(f'batch_idx: {batch_idx} \tTraining Loss: {train_loss:.6f}', end='\r')
######################
# validate the model #
######################
model.eval()
for batch_idx, (data, target) in enumerate(loaders['valid']):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
## update the average validation loss
# forward pass: compute predicted outputs by passing inputs to the model
output = model(data)
# calculate the batch loss
loss = criterion(output, target)
# update average validation loss
valid_loss = valid_loss + ((1 / (batch_idx + 1)) * (loss.data - valid_loss))
print(f'batch_idx: {batch_idx} \tValidation Loss: {valid_loss:.6f}', end='\r')
# print training/validation statistics
print('Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f} \tElapsed Time: {}'.format(
epoch,
train_loss,
valid_loss,
str(datetime.now() - start_datetime)
))
## TODO: save the model if validation loss has decreased
if valid_loss <= valid_loss_min:
print('Validation loss decreased ({:.6f} --> {:.6f}). Saving model ...'.format(
valid_loss_min,
valid_loss
))
torch.save(model.state_dict(), save_path)
valid_loss_min = valid_loss
# return trained model
return model
# train the model
model_scratch = train(30, loaders_scratch, model_scratch, optimizer_scratch,
criterion_scratch, use_cuda, 'model_scratch.pt')
# load the model that got the best validation accuracy
model_scratch.load_state_dict(torch.load('model_scratch.pt'))
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 10%.
def test(loaders, model, criterion, use_cuda):
# monitor test loss and accuracy
test_loss = 0.
correct = 0.
total = 0.
model.eval()
for batch_idx, (data, target) in enumerate(loaders['test']):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
# forward pass: compute predicted outputs by passing inputs to the model
output = model(data)
# calculate the loss
loss = criterion(output, target)
# update average test loss
test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
# convert output probabilities to predicted class
pred = output.data.max(1, keepdim=True)[1]
# compare predictions to true label
correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
total += data.size(0)
print('Test Loss: {:.6f}\n'.format(test_loss))
print('\nTest Accuracy: %2d%% (%2d/%2d)' % (
100. * correct / total, correct, total))
# call test function
test(loaders_scratch, model_scratch, criterion_scratch, use_cuda)
Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶
You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.
(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset¶
Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dogImages/train, dogImages/valid, and dogImages/test, respectively).
If you like, you are welcome to use the same data loaders from the previous step, when you created a CNN from scratch.
## TODO: Specify data loaders
import torch
import torchvision.transforms as transforms
from torchvision import datasets
imsize = 224
loader = transforms.Compose([
transforms.Resize(255),
transforms.CenterCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]
)
])
trainset = datasets.ImageFolder('/data/dog_images/train/', transform=loader)
validset = datasets.ImageFolder('/data/dog_images/valid/', transform=loader)
testset = datasets.ImageFolder('/data/dog_images/test/', transform=loader)
loaders_transfer = {}
loaders_transfer['train'] = torch.utils.data.DataLoader(trainset, batch_size=32, shuffle=True)
loaders_transfer['valid'] = torch.utils.data.DataLoader(validset, batch_size=32, shuffle=True)
loaders_transfer['test'] = torch.utils.data.DataLoader(testset, batch_size=32, shuffle=True)
(IMPLEMENTATION) Model Architecture¶
Use transfer learning to create a CNN to classify dog breed. Use the code cell below, and save your initialized model as the variable model_transfer.
import torchvision.models as models
import torch.nn as nn
## TODO: Specify model architecture
model_transfer = models.densenet121(pretrained=True)
# Freeze parameters so we don't backprop through them
for param in model_transfer.parameters():
param.requires_grad = False
model_transfer.classifier = nn.Sequential(
nn.Dropout(0.2),
nn.Linear(1024, 512),
nn.ReLU(),
nn.Dropout(0.2),
nn.Linear(512, 133)
)
if use_cuda:
model_transfer.cuda()
Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.
Answer:
The model is built using the DenseNet-121 pretrained model. In most CNNs only successive layers are connected to each other, however in the DenseNet models the layers also have access to all preceding layers. This allows the latter layers to have access to both high level and low features which improves the performance of the model as shown in the figure below.
The DenseNet models have the lowest Top-1 and Top-5 errors when benchmarked against similar pre-trained models. Ref: https://pytorch.org/docs/0.3.0/torchvision/models.html
print(model_transfer)
The classifier of DenseNet-121 is replaced by one with the following architecture:
| Name | Type | in_channels | out_channels | p | Size | Depth |
|---|---|---|---|---|---|---|
| dropout | Dropout | - | - | 0.2 | 1024 | 1 |
| fc1 | Linear | 1024 | 512 | - | 512 | 1 |
| - | ReLU | 512 | 512 | - | 512 | 1 |
| fc2 | Linear | 512 | 133 | - | 133 | 1 |
- The classifier has two fully connected layers, where the final layer has 133 nodes which are the same as the number of classes
- The first fully connected layer has size 1,024 which is the dimension of the features section of DenseNet-121
- In the fully connected layers, 20% of the hidden nodes are removed for each training pass to ensure the model doesn't become too biased towards certain hidden nodes and the the model generalizes well
Note, The normalisation of the images that are being trained must be matched with those of pre-trained model:
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
(IMPLEMENTATION) Specify Loss Function and Optimizer¶
Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_transfer, and the optimizer as optimizer_transfer below.
criterion_transfer = nn.CrossEntropyLoss()
optimizer_transfer = optim.Adam(model_transfer.classifier.parameters(), lr=0.001)
(IMPLEMENTATION) Train and Validate the Model¶
Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_transfer.pt'.
# train the model
model_transfer = train(20, loaders_transfer, model_transfer, optimizer_transfer, criterion_transfer, use_cuda, 'model_transfer.pt')
# load the model that got the best validation accuracy (uncomment the line below)
model_transfer.load_state_dict(torch.load('model_transfer.pt'))
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 60%.
test(loaders_transfer, model_transfer, criterion_transfer, use_cuda)
(IMPLEMENTATION) Predict Dog Breed with the Model¶
Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan hound, etc) that is predicted by your model.
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.
# list of class names by index, i.e. a name can be accessed like class_names[0]
class_names = [item[4:].replace("_", " ") for item in trainset.classes]
def predict_breed_transfer(img_path):
# load the image and return the predicted breed
image = image_loader(img_path)
if use_cuda:
image = image.cuda()
output = model_transfer(image)
_, preds_tensor = torch.max(output, 1)
preds = np.squeeze(preds_tensor)
return class_names[preds]
test_dog_image = dog_files[2000]
pred = predict_breed_transfer(test_dog_image)
print('Filename:', test_dog_image)
print('Predicted Breed:', pred)
img = cv2.imread(test_dog_image)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
Step 5: Write your Algorithm¶
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
- if a dog is detected in the image, return the predicted breed.
- if a human is detected in the image, return the resembling dog breed.
- if neither is detected in the image, provide output that indicates an error.
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and human_detector functions developed above. You are required to use your CNN from Step 4 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
(IMPLEMENTATION) Write your Algorithm¶
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.
def run_app(img_path):
img = cv2.imread(img_path)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
if dog_detector(img_path):
print('Found a dog in the image')
pred = predict_breed_transfer(img_path)
print('The dog breed is:', pred)
elif face_detector(img_path):
print("Didn't find a dog but found a person in the image")
pred = predict_breed_transfer(img_path)
print('This person looks like:', pred)
else:
print('Did not find a dog or a person in the image')
Step 6: Test Your Algorithm¶
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer: (Three possible points for improvement)
- The model performs well overall and above expectations considering the short amount of time that was taken to develop the project. This is doe to the power of transfer learning and leveraging pre-trained models.
- The accuracy of the
face_detector()model is good whilst the accuracy of thedog_detector()model is fantastic, showing that the performance of the VGG16 model is superior to the OpenCV impelementation. To improve the project further the following can be performed:
- The face_detector() function uses OpenCV to identify whether there is a person in the image, this takes a while to run so investigate using alternative ways of human detection, e.g. YoloV3
- Try and improve the accuracy of the CNN model which used the pre-trained model. This could be achieved by:
- Increasing the size of the training set with new dog breed images
- Performing further rotations to existing images and adding to the training set, although care needs to be taken that the model doesn't become biased towards the images in the training set
- With further images, there is scope to perform further training on the DensNet121 model such that the weights in the features part of the network are tuned towards predicting dog breeds rather than the 1000 different classes in the ImageNet database
- A further improvement to the app would be to show bounding boxes around the identified humans or dogs, and then generating dog breed predictions for each of the humans and/or dogs in the image.
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.
## suggested code, below
for file in np.hstack((human_files[:3], dog_files[:3])):
run_app(file)
Load dependencies for running app¶
import cv2
import matplotlib.pyplot as plt
from PIL import Image
import torchvision.transforms as transforms
from torch.autograd import Variable
import torch
import torchvision.models as models
import torch.nn as nn
import numpy as np
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
VGG16 = models.vgg16(pretrained=True)
use_cuda = torch.cuda.is_available()
if use_cuda:
VGG16 = VGG16.cuda()
def image_loader(image_name, imsize=224):
"""
load single image, returns tensor
Ref: https://discuss.pytorch.org/t/how-to-classify-single-image-using-loaded-net/1411
"""
loader = transforms.Compose([
transforms.Resize(255),
transforms.CenterCrop(imsize),
# transforms.RandomResizedCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])])
image = Image.open(image_name)
image = loader(image).float()
image = Variable(image, requires_grad=True)
image = image.unsqueeze(0) #this is for VGG, may not be needed for ResNet
return image
def VGG16_predict(img_path):
'''
Use pre-trained VGG-16 model to obtain index corresponding to
predicted ImageNet class for image at specified path
Args:
img_path: path to an image
Returns:
Index corresponding to VGG-16 model's prediction
'''
## TODO: Complete the function.
## Load and pre-process an image from the given img_path
## Return the *index* of the predicted class for that image
image = image_loader(img_path)
if use_cuda:
image = image.cuda()
output = VGG16(image)
_, preds_tensor = torch.max(output, 1)
preds = np.squeeze(preds_tensor)
return preds
def dog_detector(img_path):
## TODO: Complete the function.
pred = VGG16_predict(img_path)
if use_cuda:
pred = pred.cpu()
if pred.numpy() in list(range(151, 268+1)):
return True
else:
return False
model_transfer = models.densenet121(pretrained=True)
for param in model_transfer.parameters():
param.requires_grad = False
model_transfer.classifier = nn.Sequential(
nn.Dropout(p=0.2),
nn.Linear(1024, 512),
nn.ReLU(),
nn.Dropout(0.2),
nn.Linear(512, 133)
)
if use_cuda:
model_transfer.cuda()
model_transfer.load_state_dict(
torch.load('model_transfer.pt', map_location=lambda storage, loc: storage),
strict=False
)
model_transfer.eval()
import torch
import torchvision.models as models
from PIL import Image
import torchvision.transforms as transforms
from torchvision import datasets
imsize = 224
train_loader = transforms.Compose([
transforms.Resize(255),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(10),
transforms.CenterCrop(imsize),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.5, 0.5, 0.5],
std=[0.5, 0.5, 0.5]
)
])
trainset = datasets.ImageFolder('/data/dog_images/train/', transform=train_loader)
class_names = [item[4:].replace("_", " ") for item in trainset.classes]
def predict_breed_transfer(img_path):
# load the image and return the predicted breed
image = image_loader(img_path)
if use_cuda:
image = image.cuda()
output = model_transfer(image)
_, preds_tensor = torch.max(output, 1)
preds = np.squeeze(preds_tensor)
ps = torch.exp(output) / torch.sum(torch.exp(output))
res, ind = ps.topk(5)
for r, i in zip(res.view(-1), ind.view(-1)):
print(f'Breed {class_names[i]} with probability {int(r*100)}%')
return class_names[preds]
def run_app(img_path):
img = cv2.imread(img_path)
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.imshow(cv_rgb)
plt.show()
if dog_detector(img_path):
print('Found a dog in the image')
pred = predict_breed_transfer(img_path)
print('The dog breed is:', pred)
elif face_detector(img_path):
print("Didn't find a dog but found a person in the image")
pred = predict_breed_transfer(img_path)
print('This person looks like:', pred)
else:
print('Did not find a dog or a person in the image')
file_list = [
'app_test_images/trump.jpg',
'app_test_images/teddy.jpg',
'app_test_images/teddy_puppy.jpg',
'app_test_images/arlo.jpg',
'app_test_images/pk.jpg',
'app_test_images/dk.jpg',
]
for file in file_list:
run_app(file)
